In the Andrew cell we also backup 20,000 volumes nightly. Our backup
system does something like a vos backup just before it dumps the data,
but it has been modified to limit the number of writes to the VLDB (we
don't lock the VLDB entry if it already exists). The VLDB writes were
killing our DB servers before. Because of the vos backup at the time of
the dump the backup is of more recent data. The vos backup certainly is
not our bottleneck. We wait much longer on dumps of large volumes which
we do in parrallel from different servers to the local disk of the
backup machine. In our environment one of the most deadly situations is
a vos listvldb, Vos listvldb gets much better with 3.3, but still not
quite livable.
We do dumps from 7pm to the wee hours of the morning and continue to
write the data to tape all day. We have 26 fileservers (3 of these are
only DB servers - and soon to be 5 DB servers). Our current disk space
is:
Disk space used: 57584121
Disk space Available: 26422048
Disk Space Total: 84006169
